[time-nuts] Re: Timestamping counter techniques : phase computation validation question

[Erik Kaashoek]

To validate the use of regression for phase calculation an 
implementation has been made.
A single OCXO was used as input to a PLL outputting 10 MHz and 
10.0000001MHz to create two signals with a phase difference that wraps 
every 10 seconds
These two signals where send to two implementations of the counter, both 
using the same reference clock at 213.333 MHz and a capture interval of 
10 micro seconds and the difference between the measured phase of the 
two signals was recorded with Timelab.
One measurement was done without interpolation using only the event and 
clock counters at a 0.1 s measurement interval. The second measurement 
was done using interpolation of the captured counters during the 0.1 s 
measurement interval.
As a reference the same phase difference measurement was done with a 
Picotest U6200A.
The measured phase was unwrapped and the global linear trend removed and 
saved as attached tim files
The Timelab plots of the phase difference and the ADEV phase are also 
attached.

The ADEV plots of the counter show using regression only improved the 
ADEV with a factor of 10, which is less than expected from the 10000 
datapoints used in the interpolation but the ADEV measurement with 
regression shows some oscillations.  Looking at the phase difference 
plot there is a regular pattern that may be caused by leakage between 
the front-ends of the two counters probably causing the ADEV to be worse.
The plot of the frequency difference without regression (not attached) 
shows the expected quantization due to the 213.333 MHz clock.
It was not possible to set the phase/time measurement time of the 
Picotest U6200A. Timelab estimated it to be 0.02 s but in reality it 
seems more like 0.01 s as the original wrapped phase as measured by the 
Picotest wrapped about twice as slow as my own implementation.

Questions:
Is this a good way to verify the phase measurement performance?
How can the test setup further be improved?






-------------- next part --------------
A non-text attachment was scrubbed...
Name: ADEV_phase.png
Type: image/png
Size: 70391 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220201/58622d2d/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: phase_difference.png
Type: image/png
Size: 145152 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220201/58622d2d/attachment-0001.png>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: no_regression.tim
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220201/58622d2d/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: regression.tim
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220201/58622d2d/attachment-0001.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: reference.tim
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220201/58622d2d/attachment-0002.ksh>